This week’s news from Zed, Anthropic, and OpenRouter shows why better harnesses matter more than better models
thenewstack.io·13h ago
AI Research
arXiv CS.AI·Jun 9·5 min read
Decision-Aware Memory Cards: Counterfactual-Inspired Context Selection and Compression for Tool-Using LLM Agents
TL;DR
Tool-using large language model (LLM) agents often struggle to select and present relevant evidence at critical moments. CICL, a decision-aware context layer, was developed to create a context graph that organizes evidence and routes judgments through a structured schema.
✦ Why It Matters
Engineers can enhance LLM performance by implementing decision-aware context selection techniques like CICL.
Key Takeaways
- Utilize CICL for better context selection in LLMs.
- Expect significant performance improvements in retrieval tasks.
- Evaluate context based on action shift and outcome uplift.
- Be aware of performance drops when removing key context units.
How It Works
CICL operates by converting instance evidence into a context graph, which organizes information into a structured format. It routes judgments through a shared schema, scoring them based on various criteria such as necessity and risk.
This allows for effective selection and compression of high-utility evidence, packaged as memory cards for efficient retrieval.